Home / MCP / Browserbase MCP Server
Provides cloud browser automation for LLMs, enabling web interactions, data extraction, and screenshots via Browserbase and Stagehand
Configuration
View docs{
"mcpServers": {
"browserbase_http": {
"url": "your-smithery-url.com"
}
}
}The Browserbase MCP Server enables LLM applications to automate real web tasks by connecting to cloud browsers, capturing data, and performing precise actions on web pages. It combines Browserbase with Stagehand to provide fast, reliable browser automation, data extraction, and screenshots for AI workflows.
You use this MCP server by linking it to your MCP client through either a remote HTTP (SHTTP) endpoint or a local STDIO setup. The HTTP method lets you point your client at a remote hosted MCP URL, while STDIO runs the server directly in your environment. You can run the server to automate browser interactions, extract data from pages, navigate flows, and capture screenshots, all while leveraging multiple models and optional experimental features for advanced capabilities.
Prerequisites: you need Node.js and npm or a container runtime if you plan to run Docker. Ensure you have network access to the required APIs and that you can install dependencies.
# Prerequisites check (example commands)
node -v
npm -v
```
```bash
# Option A: Use the remote HTTP endpoint (no local install required beyond configuration)
# See the HTTP configuration section for details
```
```bash
# Option B: Run STDIO locally via npm path (recommended for quick local testing)
# Add the Browserbase server to your MCP config JSON as shown in the snippet below
```
} ]},{The server can also be run locally via Docker or directly from source. If you prefer local execution, follow the STDIO flow described in the configuration examples and ensure you provide any required environment keys.
You can configure the server to run with a remote HTTP URL or as a local STDIO process. The HTTP path connects to a hosted MCP endpoint, while the STDIO path runs the MCP server directly in your environment. When starting with a local build, you can adjust viewport sizes, enable experimental features, and specify a model to use with Stagehand.
{
"mcpServers": {
"browserbase": {
"type": "http",
"url": "your-smithery-url.com",
"args": []
}
}
}
```"}]} ,{If you prefer the local runtime, use the following STDIO configuration example to run the server through npm. It includes the required environment variables you should populate with your actual keys.
Key capabilities include browser automation through cloud browsers, data extraction from web pages, web interactions, full-page and element screenshots, and session management. The server supports multiple models (OpenAI, Claude, Gemini, etc.), enables vision enhancements with annotated screenshots, and provides high-performance improvements through caching in Stagehand v3.
Experimental features can be enabled with the --experimental flag during startup to access cutting-edge capabilities that may still be evolving.
Built-in page context for interacting with a browser page, including navigation and element interaction.
Locator utility to find elements within a page using selectors.
Locator for navigating into nested frames within the page context.
Advanced locator to traverse shadow DOMs and complex DOM structures for precise targeting.